S07.5 · Materials, Chemicals & Mining
The $3.5-3.8T commodity core of the global chemicals market, where AI process control is cutting energy and downtime costs and codifying operator know-how.
Bulk and commodity chemicals — petrochemicals, basic inorganics and polymers — account for roughly 60-65% of the $5.7-5.9T global chemical sales market (2024, CEFIC Facts & Figures/American Chemistry Council), about $3.5-3.8T. Growth is a modest 2-4% a year with GDP, and margins compressed materially through 2023-25 under elevated European energy costs and Chinese overcapacity together. What AI has changed so far is production economics and little else: advanced process control is cutting the two largest controllable cost lines — energy/feedstock waste and unplanned downtime — while product design and market structure sit where they were.
China accounts for more than 45% of global chemical output. The US Gulf Coast and the Middle East hold a structural feedstock-cost advantage on cheap natural gas and associated liquids; European capacity has been closing since the energy-price shock of 2022. Globally the industry is fragmented, regionally it is oligopolistic — BASF, Dow, SABIC, Sinopec and LyondellBasell lead. The chain runs from upstream oil, gas and mineral feedstock into downstream polymers, fibers, construction materials and packaging.
This is price-taking in its purest form: feedstock-cost-plus, 5-12% EBITDA at mid-cycle, swinging 2-3x between trough and peak. Thin margins even at the top of the cycle, compounded by that volatility, are the central reason a number of large diversified chemical producers have separated their commodity and specialty businesses into distinct entities in recent years — each run and valued against its own economics rather than blended together.
Advanced process control — AI systems continuously optimizing plant operating parameters — is already deployed at integrated production sites and should reach mid-market adoption within 2-5 years. Its targets are energy and feedstock waste and unplanned downtime, the two largest controllable cost lines in commodity chemical production. In a business earning single-digit to low-teens EBITDA, even modest percentage-point gains on those lines are the difference between an acceptable cycle and a bad one.
The quieter effect is on institutional knowledge, and it cuts against tenure. Process-control systems are codifying decades of operator "tribal knowledge" — the accumulated, often undocumented judgment experienced plant operators use to run a facility efficiently — into models deployable at additional sites. The moat that erodes belongs specifically to producers whose edge was the longest-tenured, most experienced operating teams: a newer entrant, or a newly acquired facility, can now approach veteran-level operating efficiency faster by importing an AI control system than by growing operator experience organically over years. Where competitive position rested heavily on operational tenure rather than feedstock access or plant scale, that advantage is narrowing.